Professional Documents
Culture Documents
Zhendong Luo
School of Mathematics and Physics
North China Electric Power University
Beijing, China
Goong Chen
Department of Mathematics
Texas A&M University
College Station, TX, USA
Series Editor
Goong Chen
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1650, San Diego, CA 92101, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
Copyright © 2019 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopying, recording, or any information storage and retrieval system, without
permission in writing from the publisher. Details on how to seek permission, further information about
the Publisher’s permissions policies and our arrangements with organizations such as the Copyright
Clearance Center and the Copyright Licensing Agency, can be found at our website:
www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher
(other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical treatment
may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and
using any information, methods, compounds, or experiments described herein. In using such information
or methods they should be mindful of their own safety and the safety of others, including parties for
whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any
liability for any injury and/or damage to persons or property as a matter of products liability, negligence
or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in
the material herein.
ISBN: 978-0-12-816798-4
We are living in an era of internet, WiFi, mobile apps, Facebook, Twitters, In-
stagram, selfies, clouds, . . . – surely we have not named them all, but one thing
certain is that these all deal with digital and computer-generated data. There is
a trendy name for the virtual space of all these things together: big data. The
size of the big data space is ever-growing with an exponential rate. Major issues
such as analytics, effective processing, storage, mining, prediction, visualiza-
tion, compression, and encryption/decryption of big data have become problems
of major interest in contemporary technology.
This book aims at treating numerical methods for partial differential equa-
tions (PDEs) in science and engineering. Applied mathematicians, scientists,
and engineers are always dealing with big data, all the time. Where do their big
data originate? They come, mostly, from problems and solutions of equations of
physical and technological systems. A large number of the modeling equations
are PDEs. Therefore, effective methods and algorithms for processing and re-
solving such data are much in demand. This is not a treatise on general big data.
However, the main objective is to develop a methodology that can effectively
help resolve the challenges of dealing with large data sets and with speedup in
treating time-dependent PDEs.
Our approach here is not the way in which standard textbooks on numerical
PDEs are written. The central theme of this book is actually on the technical
treatment for effective methods that can generate numerical solutions for time-
dependent PDEs involving only a small set of data, but yield decent solutions
that are accurate and suitable for applications. It reduces data storage, CPU time,
and, especially, computational complexity – several orders of magnitude. The
key idea and methodology is proper orthogonal decomposition (POD), from
properties of eigensolutions to a problem involving a large data set. (Indeed,
POD has been known as an effective method for big data even before the term
big data was coined.) In the process, we have developed the necessary mathe-
matical methods and techniques adapting POD to fundamental numerical PDE
methods of finite differences, finite elements, and finite volume in connection
with various numerical schemes for a wide class of time-dependent PDEs as
showcases.
xi
xii Foreword and Introduction
lighten the computational load, and reduce the running CPU time and the accu-
mulation of truncation errors in the computational processes. This approach is
based on cost-optimal, rational, and mathematical thinking and will be the one
taken by us here.
The focal topic of the book, the POD method (see [56,60]), is one of the
most effective methods that aims exactly at helping computational PDEs.
POD in a Nutshell
The POD method essentially provides an orthogonal basis for representing a
given set of data in a certain least-squares optimal sense, i.e., it offers ways to
find optimal lower-dimensional approximations for the given data set.
formulation of the POD basis, used them to establish the POD-based reduced-
order models, and then recomputed the numerical solutions on the same time
span [0, T ]. This produces some repetitive computation but not much extra gain.
One begins to ponder how to improve this and, furthermore, how to generalize
the methodology initiated by Kunisch–Volkweind’s work beyond the Galerkin
FE method to other FE methods and also to FD and FVE schemes. This book
aims exactly at answering these questions.
the classical FE method, and the mixed FE (MFE) method. Then we describe
the construction, theoretical analysis, and implementations of algorithms for the
PODROEFE methods for the following 2D PDEs: the viscoelastic wave equa-
tion, the Burgers equation, and the nonstationary parabolized Navier–Stokes
equation (for which the stabilized Crank–Nicolson extrapolation scheme is
used), respectively. Numerical examples and graphics are again illustrated.
Chapter 3 contains three sections, aiming at the treatment of PODROEFVE.
We first introduce the basics of FVE. Then three sections for the construction,
theoretical error analysis, and the implementations of algorithms for the PO-
DROEFVE methods for the following three 2D dynamic PDEs are studied: the
hyperbolic equation, Sobolev equation, and incompressible Boussinesq equa-
tion, respectively, are developed, with concrete examples and illustrations.
Numerical results on these model equations as presented in the book have
demonstrated the effectiveness and accuracy of our POD methods.
Finally, Chapter 4 is a short epilogue and outlook, consisting of concluding
remarks and forward-looking statements.
The book is written to be as self-contained as possible. Readers and students
need only to have an undergraduate level in applied and numerical mathemat-
ics for the understanding of this book. Many parts can be used as a standard
graduate-level textbook on numerical PDEs. The theory, methods, and com-
putational algorithms will be valuable to students and practitioners in science,
engineering, and technology.
ACKNOWLEDGMENTS
The authors thank all collaborators, colleagues, and institutions that have gen-
erously supported our work. In particular, the authors are delighted to acknowl-
edge the partial financial support over the years by the National Natural Science
Foundation of China (under grant #11671106), the Qatar National Research
Fund (under grant #NPRP 8-028-1-001), the North China Electric Power Uni-
versity, and the Texas A&M University.
Zhendong Luo
Beijing, China
Goong Chen
College Station, TX, USA
Chapter 1
Reduced-Order Extrapolation
Finite Difference Schemes
Based on Proper Orthogonal
Decomposition
The key objective of this book is to develop the numerical treatments of
proper orthogonal decomposition (POD) for partial differential equations
(PDEs). With regard to numerical methods for PDEs, the finite difference
(FD) method essentially constitutes the basis of all numerical methods for
PDEs. In order to introduce how POD works, it is natural to start with the
FD method.
For the sake of proper self-containedness, in this chapter we first review the
basic theory of classical FD schemes. We then introduce the construction, theo-
retical analysis, and implementations of algorithms for the POD-based reduced-
order extrapolation FD (PODROEFD) schemes for the two-dimensional (2D)
parabolic equation, 2D nonstationary Stokes equation, and 2D shallow wa-
ter equation with sediment concentration. Finally, we provide some numerical
examples to show what the PODROEFD schemes have over the classical FD
schemes. Moreover, it is shown that the PODROEFD schemes are reliable and
effective for solving above-mentioned PDEs.
The numerical models treated here include both simple equations and cou-
pled systems. The systematic approach we take in this chapter, namely, fol-
lowing the logical sequence of rudiments, modeling equations, error estimates-
stability-convergence, POD methods, error estimates for POD solutions, and
finally concrete numerical examples, will be the standard for all chapters.
1
μx unj = (uj + 1 + uj − 1 ); μx is known as the average operator and satisfies
2 2 2
1 12 − 12
μx = (Ex + Ex ).
2
2. Composite difference operators
We have
1 1
i. (μδ)x = (Ex − Ex−1 ) = (x + ∇x );
2 2
1
− 12 2
ii. (δx ) = δx δx = (Ex − Ex ) = (Ex − 2I + Ex−1 );
2 2
un+1
j − unj δ
= (un − 2unj + unj−1 )
t x 2 j +1
∂u ∂ 2u
is an explicit FD scheme for = δ 2 , where δ is a positive “diffusion coef-
∂t ∂x
ficient”.
2. Implicit FD schemes
The implicit FD schemes imply that the time farthest point values in the FD
scheme appear more than once. For example,
un+1
j − unj δ
= (un+1 − 2un+1 + un+1
j −1 )
t x 2 j +1 j
∂u ∂ 2u
is an implicit FD scheme for =δ 2.
∂t ∂x
3. Semidiscretized difference schemes
The semidiscretized difference scheme implies that only the spatial variable
is discretized and the time variable is not discretized. For example,
du δ
= (uj +1 − 2uj + uj −1 )
dt j x 2
∂u ∂ 2u
is a semidiscretized difference scheme for =δ 2.
∂t ∂x
un+1
j − unj δ
= (un − 2unj + unj−1 ),
t x 2 j +1
Reduced-Order Extrapolation Finite Difference Schemes Chapter | 1 5
The above equation is known as the modified PDE, where the first parenthesis
above is called the source equation (i.e., PDE) and the bracket above is called
the remainder (R) or truncation error (TE), denoted by R = TE.
The TE is equal to the source equation (PDE) subtracting the FD equation
(FDE), i.e., TE = PDE − FDE.
The discretization error (DE, i.e., the error of the FD scheme) is equal to
the TE plus the boundary error (BE), i.e., DE = TE + BE. Thus, DE = BE +
PDE − FDE.
The round-off error (ROE) denotes the rounding error of the computing pro-
cedure by the computer and the total error is denoted by CE. Then
However, ROE and BE are usually omitted. The error of FD schemes mainly
considers TE, which is directly obtained from the approximation of the deriva-
tive (1.1.1)–(1.1.4).
2. The consistency of an FD scheme
ut = Lu (1.1.5)
Definition 1.1.2. i. If an error disturbance εjn = ũnj − unj is added at certain time
level t = tn , i.e., ũnj = unj + εjn , and the errors εjn+1 = ũn+1
j − un+1
j of solutions
ũn+1
j and un+1
j obtained from the FD scheme (1.1.6) do not produce extra-large
overall growth, i.e., there is a constant K > 0 independent of n and j such that
where · represents a norm, then when 0 < K < 1, the FD scheme (1.1.6) is
said to be strongly stable, else (when K 1) the FD scheme (1.1.6) is said to be
weakly stable.
ii. If there is no restriction between the time step and the spatial step in the
FD scheme (1.1.6), it is said to be absolutely stable or unconditionally stable.
In this case, the time step t can take a larger size.
iii. If the stability of the FD scheme (1.1.6) is restricted by some relationship
of time step and spatial steps (usually constrained in the form of some inequal-
ities, for example, t = o(x γ ) (γ > 0)), then it is said to be conditionally
stable.
We have the following criteria for the stability of the FD scheme (1.1.6) (see
[192,194]).
Theorem 1.1.1. The FD scheme (1.1.6) is stable if and only if there are two
positive constants t0 and x0 as well as a nonnegative constant K > 0 inde-
pendent of n and j such that the solutions of FD scheme (1.1.6) satisfy
unj K, n = 1, 2, . . . .
un+1
j
un+1 · unj > 0, < 1, n = 1, 2, · · · , (1.1.8)
j
unj
Theorem 1.1.3. If the PDE (1.1.5) is well posed, i.e., it has a unique solution
and is continuously dependent on initial boundary value conditions and com-
patible with its FD scheme (1.1.6), then the stability of the FD scheme (1.1.6) is
equivalent to its convergence.
Theorem 1.1.3 is briefly described as “if a PDE is well posed and compatible
with its FD scheme, then the stability of the FD scheme (1.1.6) is equivalent to
its convergence”.
Remark 1.1.1. For linear PDEs, if their FD schemes are stable, then their con-
vergence is ensured. Therefore, it is only necessary to discuss their stability.
However, for nonlinear PDEs, because of the complexity, at present we can
only investigate nearly similar convergence and local stability analysis instead
of global analysis of convergence.
is said to be well posed, if it has a unique and stable solution. A so-called stable
solution means that the solution is continuously dependent on the initial value
and can maintain the boundedness of small disturbances.
Assume that the initial value φ(x) is periodic and expandable into the fol-
lowing Fourier series:
φ(x) = fk eikx , (1.1.11)
k
where the fk s are the Fourier coefficients. Assume that the exact solution u(x, t)
of the initial value problem (1.1.10) is also expandable into the following Fourier
series:
u(x, t) = Fk (t)eikx , (1.1.12)
k
where the Fk (t)s are the Fourier coefficients of the series depending on t.
8 Proper Orthogonal Decomposition Methods for Partial Differential Equations
By comparing the coefficients eikx of the LHS and those of the RHS in (1.1.14)
and setting them equal, we have
Note that the exact solution u(x, t) is said to be stable, if there is a nonnegative
constant M, independent of u, t, and x, such that
π 1/2
where u(x, t)L2 = −π | u(x, t) |2 dx is the norm of u(·, t) in L2 (−π, π).
Reduced-Order Extrapolation Finite Difference Schemes Chapter | 1 9
Thus, the exact solution u(x, t) of (1.1.10) is stable if and only if there is a
nonnegative constant M independent of u, t, and x, such that | eL(ik)t | M.
ii. Stability analysis of an FD scheme
In the following, we analyze the von Neumann stability conditions of an FD
scheme.
Let (1.1.10) have the following FD scheme:
αμ un+1
j +μ = βγ unj+γ . (1.1.23)
μ γ
where Z is the set of all integers and G is known as the growth factor, which is
denoted by the following equation:
−1 ⎛ ⎞
G= αμ eikμx ·⎝ βγ eikγ x ⎠ . (1.1.27)
μ γ
g n = Gg n−1 , n = 1, 2, · · · . (1.1.28)
10 Proper Orthogonal Decomposition Methods for Partial Differential Equations
Gn K, n = 1, 2, · · · . (1.1.31)
Corollary 1.1.5. The FD scheme (1.1.13) is stable if and only if its growth
factor G satisfies
into the FD scheme (1.1.13), and then, eliminating some common factors, one
can obtain the growth factor G of (1.1.27).
For an FD scheme for two-dimensional linear PDEs, we need only to substi-
tute unj,m = Gn eikxj eikym (n = 1, 2, · · · ) into the FD scheme and then simplify,
so we can also obtain the growth factor G.
Some relative examples can be found in [192].
where f (x, y, t), g(x, y, t), and s(x, y) are given source item, boundary func-
tion, and initial value function, respectively, and T is the total time duration.
The main motivation and physical background of the parabolic equation are
the modeling of heat conduction and diffusion phenomena.
which shows that solutions of the FD scheme (1.2.2) are bounded and contin-
uously dependent on the initial value s(x, y) and source term f (x, y, t). Thus,
by Theorem 1.1.1, we deduce that the FD scheme (1.2.2) is stable. Furthermore,
the error estimates (1.2.3) are directly obtainable from approximating difference
quotients for derivatives.
Thus, as long as the time step t, the spatial steps x and y, f (x, y, t),
g(x, y, t), and s(x, y) are given, we can obtain classical FD solutions unj,k (0
j J, 0 k K, 0 n N ) for the 2D parabolic equation (1.2.1) by comput-
ing the FD scheme (1.2.2).
Because the number of mesh points m is much larger than that of snap-
shots L extracted, the height m of A for matrix AAT is much larger than the
width L of A for matrix AT A, but the positive eigenvalues λj (j = 1, 2, · · · , l)
of AT A and AAT are identical and satisfy λj = σj2 (j = 1, 2, · · · , l). There-
fore, we may first find the eigenvalues λ1 λ2 · · · λl > 0 (l = rank A) for
matrices AT A and corresponding eigenvectors ϕ j . Then, by the relationship
ϕ j = Aφ j / λj , j = 1, 2, . . . , l,
M
AM = σi ϕ j φ Ti = T A. (1.2.11)
i=1
Proof. According to
⎛ ⎞⎛ ⎞
σ1 ··· 0 φ T1
M
⎜ . ⎟⎜ ⎟
AM = σi ϕ i φ Ti = ϕ1 · · · ϕM ⎜ . .. .. ⎟ ⎜ .. ⎟,
⎝ . . . ⎠⎝ . ⎠
i=1
0 ··· σM φ TM
⎛ ⎞⎛ ⎞
σ1 ··· 0 φ T1
⎜ . ⎟⎜ ⎟
A= ϕ1 · · · ϕl ⎜⎝ ..
..
.
..
.
⎟ ⎜ ..
⎠⎝ .
⎟,
⎠
0 ··· σl φ Tl
we have
14 Proper Orthogonal Decomposition Methods for Partial Differential Equations
⎛ ⎞ ⎛ ⎞⎛ ⎞
ϕ T1 σ1 · · · 0 φ T1
⎜ ⎟ ⎜ . ⎟⎜ ⎟
A = ⎜
T . ⎟
⎝ .. ⎠ ϕ 1 · · · ϕ l ⎝ ..
⎜ ..
.
.. ⎟ ⎜ ..
. ⎠⎝ .
⎟
⎠
T
ϕM 0 ··· σl φ Tl
⎛ ⎞⎛ ⎞
σ1 · · · 0 φ T1
⎜ . ⎟⎜ ⎟
= IM O ⎜ ⎝ ..
.. . ⎟⎜ . ⎟
. .. ⎠ ⎝ .. ⎠
0 · · · σl φ Tl
⎛ ⎞⎛ ⎞
σ1 · · · 0 · · · 0 φ T1
⎜ ⎟⎜ ⎟
= ⎜ . ..
⎝ .. .
..
.
. ⎟⎜ . ⎟
· · · .. ⎠ ⎝ .. ⎠
0 · · · σM · · · 0 φ Tl
⎛ ⎞⎛ ⎞
σ1 · · · 0 φ T1
⎜ . ⎟⎜ ⎟
= ϕ1 · · · ϕM ⎜ ⎝ ..
..
.
.. ⎟ ⎜ ..
. ⎠⎝ .
⎟ = AM .
⎠
0 · · · σM φT M
Thus, by the relationship between the matrix norm and its spectral radius,
we have
min A − B2,2 = A − AM 2,2 = A − T A2,2 = λM+1 ,
rank(B)M
(1.2.12)
the unit vector with the nth component being 1. The inequality (1.2.13) √ shows
that unM is an optimal approximation of un whose error is no more than λM+1 .
Thus, is just an orthonormal optimal POD base of A.
⎛ ⎞
J times 0
⎜ −1 ⎟
⎜ ··· ··· 1 ⎟
⎜ ⎟
⎜ .. ⎟
⎜ −2 . ⎟
⎜ ⎟
⎜ .. ⎟
⎜ −2 . ⎟
⎜ ⎟
⎜ ⎟
C=⎜
⎜ 1
.. ⎟ .
⎜
. 1 ⎟
⎟
⎜ ⎟
⎜ .. ⎟
⎜ . −2 ⎟
⎜ ⎟
⎜ .. ⎟
⎜ . −2 ⎟
⎜ ⎟
⎝ 1 ·· · · ·· −1 ⎠
J times 0 m×m
Therefore, by using the relationship between the matrix eigenvalue and its
norm, we have
⎪ α n+1 = α n + x t
2 Bα + y 2 Cα + t F m ,
T n t T n T n (1.2.17)
⎪
⎩
n = L, L + 1, · · · , N − 1.
u∗n = α n , n = 1, 2, · · · , L, L + 1, · · · , N. (1.2.18)
Further, we can obtain the PODROEFD solution components for Eq. (1.2.1)
as follows:
u∗n ∗n
j,k = ui , 0 j J, 0 k K,
1 i = k(J + 1) + j + 1 m = (K + 1)(J + 1). (1.2.19)
Remark 1.2.1. It is easily seen that the classical FD scheme (1.2.2) at each
time level contains m unknown quantities, whereas the system of Eqs. (1.2.17)–
(1.2.18) at the same time level (when n > L) contains only M unknown quanti-
ties (M L m, usually M = 6, but m = O(104 ) ∼ O(106 )). Therefore, the
PODROEFD model (1.2.17)–(1.2.18) includes very few degrees of freedom and
does not involve repeated computations. Here, we extract the snapshots from the
Reduced-Order Extrapolation Finite Difference Schemes Chapter | 1 17
t t
u∗n+1 = u∗n + 2
Bu∗n + Cu∗n + tF nm ,
x y 2
n = L, L + 1, · · · , N − 1. (1.2.21)
By summarizing the above discussion and noting that the absolute value of
each vector component is less than the vector norm, from (1.2.3), we obtain the
following theorem.
Theorem 1.2.4. The errors between the solutions unjk from the classical FD
scheme and the solutions u∗n j k from the PODROEFD scheme (1.2.17)–(1.2.18)
satisfy the following estimates:
| unjk − u∗n
j k | E(n) λM+1 , 1 j J, 1 k K, 1 n N, (1.2.23)
1 j J, 1 k K, 1 n N. (1.2.24)
18 Proper Orthogonal Decomposition Methods for Partial Differential Equations
√
Remark 1.2.2. The error terms containing λM+1 in Theorem 1.2.4 are
caused by the POD-based reduced-order for the classical FD scheme, which
could be used to select
√ the number of the POD basis, i.e., it is necessary to
take M such that λM+1 = O(t, x 2 , y 2 ). Whereas E(n) = (1 + δ)n−L
(L + 1 n N ) are caused by extrapolating iterations, which√could be
used as the guide for renewing the POD basis, i.e., if (1 + δ)n−L λM+1 >
max(t, x 2 , y 2 ), it is √
necessary to update the POD basis. If we take λM+1
that satisfies (1 + δ)N−L λM+1 = O(t, x 2 , y 2 ), then the PODROEFD
scheme (1.2.17)–(1.2.18) is convergent and, thus, we do not have to update the
POD basis.
BRANCH I. EUCOPEPODA.
Sub-Order 1. Gymnoplea.
The division between the front and hind part of the body falls
immediately in front of the genital openings and behind the fifth thoracic
feet. The latter in the male are modified into an asymmetrical copulatory
organ.
TRIBE I. AMPHASCANDRIA.
TRIBE II.
HETERARTHRANDRIA.
The first antennae of the male are asymmetrical, one, usually the
right, being used as a clasping organ.
The males of the Centropagidae, Candacidae and Pontellidae,
besides possessing the asymmetrically modified thoracic limbs of the
fifth pair also exhibit a modification of one of the first antennae,
which is generally thickened in the middle, and has a peculiar joint in
it, or geniculation, which enables it to be flexed and so used as a
clasping organ for holding the female.
Fam. 1.—Centropagidae.—These Copepods are very common
in the pelagic plankton, and some of the species vie with the
Calanidae in plumed ornaments, e.g. Augaptilus filigerus, figured by
Giesbrecht in his monograph. The use of these ornaments, which are
possessed by so many pelagic Copepods, is entirely obscure.[40]
Certain of the Centropagidae live in fresh water. Thus Diaptomus is
an exclusively fresh-water genus, and forms a most important
constituent of lake-plankton; various species of Heterocope occur in
the great continental lakes, and certain Eurytemora go up the
estuaries of rivers into brackish water.
An excellent work on the fresh-water Copepods of Germany has
been written by Schmeil,[41] who gives analytical tables for
distinguishing various genera and species. The three fresh-water
families are the Centropagidae, Cyclopidae, and Harpacticidae (see
p. 62). The Centropagidae may be sharply distinguished from the
other fresh-water families by the following characters:—The
cephalothorax is distinctly separated from the abdomen; the first
antennae are long and composed of 24–25 segments, in the male
only a single antenna (generally the right) being geniculated and
used as a clasping organ. The fifth pair of limbs are not rudimentary;
a heart is present, and only one egg-sac is found in the female. The
second antennae are distinctly biramous.
Diaptomus.—The furcal processes are short, at most three times as
long as broad; endopodite of the first swimming appendage 2–
jointed, endopodites of succeeding legs 3–jointed.
Heterocope.—The furcal processes are short, at most twice as long
as broad; endopodites of all swimming legs 1–jointed.
Eurytemora.—The furcal processes are long, at least three and a
half times as long as broad; the endopodite of the first pair of legs
1–jointed, those of the other pairs 2–jointed.
It has been known for a long time that some of the marine
Copepods are phosphorescent, and, indeed, owing to their numbers
in the plankton, contribute very largely to bring about that liquid
illumination which will always excite the admiration of seafarers. In
northern seas the chief phosphorescent Copepods belong to
Metridia, a genus of the Centropagidae; but in the Bay of Naples
Giesbrecht[42] states that the phosphorescent species are the
following Centropagids: Pleuromma abdominale and P. gracile,
Leuckartia flavicornis and Heterochaeta papilligera; Oncaea
conifera is also phosphorescent. It is often stated that Sapphirina (p.
69) is phosphorescent, but its wonderful iridescent blue colour is
purely due to interference
colours, and has nothing to do
with phosphorescence.
Giesbrecht has observed that the
phosphorescence is due to a
substance secreted in special
skin-glands, which is jerked into
the water, and on coming into
contact with it emits a
phosphorescent glow. This
substance can be dried up
completely in a desiccated
specimen and yet preserve its
phosphorescent properties, the
essential condition for the actual
emission of light being contact
with water. Similarly, specimens
preserved in glycerine for a long
period will phosphoresce when
compressed in distilled water.
From this last experiment
Giesbrecht concludes that the
phosphorescence can hardly be
due to an oxidation process, but
the nature of the chemical
reaction remains obscure.
Fam. 2. Candacidae.—This
family comprises the single genus
Candace, with numerous species
distributed in the plankton of all
seas. Some species, e.g. C.
pectinata, Brady, have a
practically world-wide
distribution, this species being
recorded from the Shetlands and
from the Philippines.
Fam. 3. Pontellidae.—This is
a larger family also comprising
widely distributed species found
Fig. 27.—Dorsal view of Anomalocera
pattersoni, ♂, × 20. (After Sars.)in the marine plankton.
Anomalocera pattersoni (Fig. 27)
is one of the commonest elements in the plankton of the North Sea.
Sub-Order 2. Podoplea.
The boundary between the fore and hind part of the body falls in
front of the fifth thoracic segment. The appendages of the fifth
thoracic pair in the male are never modified as copulatory organs.
TRIBE I. AMPHARTHRANDRIA.
The first antennae in the male differ greatly from those in the
female, being often geniculated and acting as prehensile organs.
Fams. 1–2. Cyclopidae and Harpacticidae, and other allied
families, are purely free-living forms; they are not usually pelagic in
habit, but prefer creeping among algae in the littoral zone or on the
sea-bottom, or especially in tidal pools. Some genera are,
nevertheless, pelagic; e.g. Oithona among Cyclopidae; Setella,
Clytemnestra, and Aegisthus among Harpacticidae.
The sketch (Fig. 28) of Euterpe acutifrons ♀ , a species widely
distributed in the Mediterranean and northern seas, exhibits the
structure of a typical Harpacticid, while Fig. 29 shows the form of the
first antenna in the male.
Several fresh-water representatives of these free-living families
occur. The genus Cyclops (Cyclopidae) is exclusively fresh-water,
while many Harpacticidae go up into brackish waters: for example
on the Norfolk Broads, Mr. Robert Gurney has taken Tachidius
brevicornis, Müller, and T. littoralis, Poppe; Ophiocamptus
brevipes, Sars; Mesochra lilljeborgi, Boeck; Laophonte littorale, T.
and A. Scott; L. mohammed, Blanchard and Richard; and
Dactylopus tisboides, Claus.
Schmeil[43] gives the following scheme for identifying the fresh-
water Cyclopidae and Harpacticidae (see diagnosis of Centropagidae
on p. 59):—
Fam. 1. Cyclopidae.—The
cephalothorax is clearly separated
from the abdomen. The first
antennae of the female when bent
back do not stretch beyond the
cephalothorax; in the male both
of them are clasping organs. The
second antennae are without an
exopodite. The fifth pair of limbs
are rudimentary, there is no
heart, and the female carries two
egg-sacs.
Cyclops.—Numerous species, split
up according to segmentation of
rudimentary fifth pair of legs,
number of joints in antennae, etc.
Fam. 2. Harpacticidae.—
The cephalothorax is not clearly
separated from the abdomen. The
first antennae are short in both
sexes, both being clasping organs
in the male. The second antennae
have a rudimentary exopodite.
The fifth pair of limbs are
rudimentary and plate-shaped; a
heart is absent, and the egg-sacs
Fig. 28.—Euterpe acutifrons, ♀ , × 70. of the female may be one or two
Abd.1, 1st abdominal segment; Th.5, in number.
5th thoracic segment. (After
Giesbrecht.) 1. Ophiocamptus (Moraria).—Body
worm-shaped; first antennae of
female 7–jointed, rostrum
forming a broad plate.
2. Body not worm-shaped; first antennae of female 8–jointed,
rostrum short and sharp.
(a) Endopodites of all thoracic limbs 3–jointed. The first
antennae in female distinctly bent after the second joint.
Nitocra.
(b) Endopodite of at least
the fourth limb 2–jointed;
first antennae in female
not bent. Canthocamptus.
3. Ectinosoma.—Body as in
2, but first antennae are very
short, and the maxillipede
does not carry a terminal
hooked seta as in 1 and 2.
Fam. 3. Peltiidae.[44]—This is
an interesting family, allied to the
Harpacticidae, and includes
species with flattened bodies
somewhat resembling Isopods,
and a similar habit of rolling
themselves up into balls. No
parasitic forms are known,
though Sunaristes paguri on the
French and Scottish coasts is said
to live commensally with hermit-
crabs.
The first antennae are short, similar in the two sexes, and are
never used by the male as clasping organs. This function may be
subserved by the second maxillae.
Fams. Oncaeidae, Corycaeidae, Lichomolgidae, Ergasilidae,
Bomolochidae, Chondracanthidae, Philichthyidae,
Nereicolidae, Hersiliidae, Caligidae, Lernaeidae,
Lernaeopodidae, Choniostomatidae.
The families Oncaeidae and Corycaeidae contain pelagic forms of
flattened shape and great swimming powers, but the structure of the
mouth-parts in the Corycaeidae points to a semi-parasitic habit.
Fam. 1. Oncaeidae.—This family, including the genera Oncaea,
Pachysoma, etc., does not possess the elaborate eyes of the next
family, nor is the sexual dimorphism so marked.
Fam. 2. Corycaeidae.—These are distinguished from the
Oncaeidae, not only by their greater beauty, but also by the
possession of very elaborate eyes, which are furnished with two
lenses, one at each end of a fairly long tube. The females of
Sapphirina are occasionally found in the branchial cavity of Salps,
and their alimentary canal never contains solid particles, but is filled
with a fluid substance perhaps derived by suction from their prey. S.
opalina may occur in large shoals, when the wonderful iridescent
blue colour of the males makes the water sparkle as it were with a
sort of diurnal phosphorescence. The animal, however, despite the
opinion of the older observers, is not truly phosphorescent. It may be
that the ornamental nature of some of the males is correlated with
the presence of the curious visual organs, which are on the whole
better developed in the females than in the males. As in so many
pelagic Copepods, the body and limbs may bear plumed setae of
great elaboration and beautiful colour, e.g. Copilia vitrea (Fig. 37).
We now pass on to the rest of the parasitic Copepods,[51] which
probably belong to the tribe Isokerandria, and we meet with the
same variety of degrees of parasitism as in the Ampharthrandria,
often leading to very similar results.
Fig. 37.—Copilia vitrea (Corycaeidae), ♀, × 20. (After Giesbrecht.)